Exploring Phylogenetic and Functional Signals in Complex Morphologies: The Hamate of Extant Anthropoids as a Test-Case Study

Page created by Elsie Mcgee
 
CONTINUE READING
Exploring Phylogenetic and Functional Signals in Complex Morphologies: The Hamate of Extant Anthropoids as a Test-Case Study
THE ANATOMICAL RECORD 298:212–229 (2015)

              Exploring Phylogenetic and Functional
                Signals in Complex Morphologies:
              The Hamate of Extant Anthropoids as a
                         Test-Case Study
                      SERGIO ALMECIJA,        1,2,3
                                                     * CALEY M. ORR,4 MATTHEW W. TOCHERI,5,6
                                    BIREN A. PATEL,7,8 AND WILLIAM L. JUNGERS1
                1
                 Department of Anatomical Sciences, Stony Brook University School of Medicine, Stony
                                                          Brook, New York
              2
                Institut Catal  a de Paleontologia Miquel Crusafont, Universitat Autonoma de Barcelona,
               Edifici Z (ICTA-ICP), campus de la UAB, c/ de les Columnes, s/n., 08193 Cerdanyola del
                                                      Vallès, Barcelona, Spain
                                                3
                                                  NYCEP Morphometrics Group
                          4
                            Department of Anatomy, Midwestern University, Downers Grove, Illinois
                 5
                   Human Origins Program, Department of Anthropology, National Museum of Natural
                  History, Smithsonian Institution, 10th and Constitution Avenue NW, Washington, DC
                6
                  Department of Anthropology, Center for the Advanced Study of Hominid Paleobiology,
                                    The George Washington University, Washington, DC
                   7
                     Cell and Neurobiology, Keck School of Medicine, University of Southern California,
                                                      Los Angeles, California
              8
                Human and Evolutionary Biology Section, Department of Biological Sciences, University
                                        of Southern California, Los Angeles, California

                                                      ABSTRACT
                           Three-dimensional geometric morphometrics (3DGM) is a powerful
                      tool for capturing and visualizing the “pure” shape of complex structures.
                      However, these shape differences are sometimes difficult to interpret
                      from a functional viewpoint, unless specific approaches (mostly based on
                      biomechanical modeling) are employed. Here, we use 3DGM to explore
                      the complex shape variation of the hamate, the disto-ulnar wrist bone, in
                      anthropoid primates. Major trends of shape variation are explored using
                      principal components analysis along with analyses of shape and size cova-
                      riation. We also evaluate the phylogenetic patterning of hamate shape by
                      plotting an anthropoid phylogenetic tree onto the shape space (i.e., phylo-
                      morphospace) and test against complete absence of phylogenetic signal
                      using posterior permutation. Finally, the covariation of hamate shape and
                      locomotor categories is explored by means of 2-block partial least squares
                      (PLS) using shape coordinates and a matrix of data on arboreal locomotor
                      behavior. Our results show that 3DGM is a valuable and versatile tool for
                      characterizing the shape of complex structures such as wrist bones in
                      anthropoids. For the hamate, a significant phylogenetic pattern is found
                      in both hamate shape and size, indicating that closely related taxa are
                      typically the most similar in hamate form. Our allometric analyses show
                      that major differences in hamate shape among taxa are not a direct con-

  Grant sponsors: Fulbright Commission and the Generalitat        *Correspondence to: Sergio Alm
                                                                                                ecija; Department of Anatomi-
de Catalunya (S.A.), the Spanish Ministerio de Economıa y      cal Sciences, Stony Brook University School of Medicine, Stony
Competitividad (S.A.), the AAPA Professional Development        Brook, NY 11794-8081. E-mail: sergio.almecija@stonybrook.edu
Grant (S.A.); the Smithsonian Scholarly Studies Grant Program     Received 3 October 2014; Accepted 11 October 2014.
(M.W.T.); Leakey Foundation (B.A.P.); Wenner-Gren Foundation
                                                                   DOI 10.1002/ar.23079
(C.M.O.); National Science Foundation; Grant numbers: 2009
                                                                Published online in Wiley Online Library (wileyonlinelibrary.
BFUL 00049, 2009 BP-A 00226, CGL2011-27343, NSF-BCS             com).
1316947, NSF-BCS-1317047, NSF-BCS 1317029.

C 2014 WILEY PERIODICALS, INC.
V
Exploring Phylogenetic and Functional Signals in Complex Morphologies: The Hamate of Extant Anthropoids as a Test-Case Study
SHAPE ANALYSIS OF EXTANT ANTHROPOID HAMATES                                           213

                        sequence of differences in hamate size. Finally, our PLS indicates a sig-
                        nificant covariation of hamate shape and different types of arboreal loco-
                        motion, highlighting the relevance of this approach in future 3DGM
                        studies seeking to capture a functional signal from complex biological
                        structures. Anat Rec, 298:212–229, 2015. VC 2014 Wiley Periodicals, Inc.

                        Key words: wrist morphology; 3D geometric morphometrics;
                                   shape–function complex

   In primates, the hamate forms the distal ulnar aspect              Lewis (1972) described the wrist morphology of cercopi-
of the carpus (Fig. 1). It articulates proximally with the         thecoids as being largely similar across major taxonomic
triquetrum (sometimes with the lunate as well), radially           groups with colobines and cercopithecines exhibiting little
with the capitate, and distally with the fourth and fifth          variation in hamate form. He described both groups as
metacarpal bases. It has a hook-shaped projection, the             exhibiting a wedge-shaped body set obliquely within the
hamulus, on its distopalmar side, which is the attachment          wrist, with a proximally oriented triquetral facet that
site for two intrinsic muscles of the fifth digit (flexor digiti   forms a stable platform. The orientation of the articular
minimi and opponens digiti minimi). Variation in hamate            facets and the corresponding blocky triquetrum and well-
shape has been used in studies of the evolution of modern          developed styloid process of the ulna probably facilitate
human manipulative capabilities because human hamate               load transfer on the ulnar side of the wrist while restrict-
morphology appears to favor the ability to form a variety          ing ulnar deviation (Lewis, 1989; Sarmiento, 1988).
of grips that are employed to construct and manipulate             O’Connor (1975) explained the relative uniformity of
tools (Marzke et al., 1992, 1998; Marzke and Marzke,               hamate morphology among cercopithecoid monkeys as
2000; Marzke, 2013; Orr et al., 2013). The functional dif-         resulting from their use of quadrupedal locomotion involv-
ferentiation of monkey and ape wrists has also been a              ing dorsiflexion of the wrist either in palmigrade or digiti-
topic of particular interest for primate morphologists             grade postures (irrespective of habitat/substrate
(Lewis, 1965, 1969, 1972; McHenry and Corrucini, 1975;             differences; see also Patel, 2009). Hyperextenison at the
O’Connor, 1975; Cartmill and Milton, 1977; Sarmiento,              fifth carpometacarpal joint (and the fourth to a lesser
1988; Lewis, 1989; Richmond, 2006). Specifically, hamate           degree) appears to facilitate such hand postures in pro-
morphology has been important in discussions of homi-              nograde quadrupeds. This is permitted by the lack of a
noid locomotor evolution because of its role in wrist mobil-       developed “hook” on the hamate (i.e., hamulus) together
ity (i.e., ranges of motion), which is necessary for               with a proximal articular surface on the fourth and fifth
behaviors such as vertical climbing and below-branch sus-          metacarpals that extends onto the dorsal surface of the
pension used by extant apes (e.g., Tuttle, 1967; Lewis,            shaft (Corruccini, 1975; O’Connor, 1975).
1972; Jenkins and Fleagle, 1975; O’Connor, 1975; Lewis,               In contrast to the minimal morphological variation in
1977; Beard et al., 1986; Sarmiento, 1988; Spoor et al.,           the hamate documented thus far in quadrupedal mon-
1991; Sarmiento, 1994; Fleagle, 1999).                             keys, considerable diversity in hamate shape has been
   Monkeys and extant hominoid primates differ consid-             reported for extant apes and humans. In both Pongo and
erably in terms of ulnar wrist morphology. Most of these           the hylobatids (that highly rely heavily on below-branch
differentiating features are related to the derived com-           suspension), the hamate is described as composing a
plex in crown hominoids in which the ulnar styloid pro-            larger proportion of the midcarpal articular surface and
cess is retracted from the proximal carpal row (Lewis,             contributing to a quasi-ball-and-socket joint that allows
1989; Sarmiento, 1988). Although the expression varies             considerable mobility (Jenkins and Fleagle, 1975; Jen-
across apes, the imposition of fibrocartilage between the          kins, 1981; Richmond et al., 2001). Lewis (1972, 1989)
ulnar styloid process and the triquetropisiform complex            also described hylobatid hamates as retaining a some-
(and reduction of the triquetrum) is thought to dramati-           what oblique-set within the overall carpus (similar to
cally alter the load transmission regime through the               monkeys). However, the hylobatid hamate has an
wrist. This reorganization has been traditionally related          ulnarly facing triquetral facet and a rounded proximal
to the capacity for increased supination at the distal             “head” that articulates with the lunate and nearly
radioulnar joint necessary for below-branch locomotion             excludes this latter bone from articulating with the capi-
(Lewis, 1972; Sarmiento, 1985, 1988; Lewis, 1989). How-            tate. The proximal aspects of the hamate and capitate
ever, the loss of ulnocarpal articulation in the 12                together contribute to a globular distal midcarpal row
million-year-old fossil great ape Pierolapithecus catalau-         that is engulfed by the lunate, scaphoid, and centrale
nicus, in combination with its inferred orthograde body            (e.g., Jenkins and Fleagle, 1975; Lewis, 1989). Conse-
plan (Moy a-Sola et al., 2004) but moderate hand length          quently, it does not present a large, flat, platform-like
(Moy a-Sol
           a et al., 2005; Almecija et al., 2009; Alba et al.,    articulation. Thus, the hylobatid configuration is in
2010), suggests that changes in the ulnar side of the              stark contrast to the stable weight-bearing platform for
wrist might have been initially related to extant great-           the triquetrum in monkeys. Lewis (1972, 1989) consid-
ape-like vertical climbing and not specifically suspen-            ered the hylobatid midcarpus to be a less well-developed
sion. It is likely that differences in hamate shape                version of the great ape wrist, which he viewed as
between apes and monkeys at least in part reflect that             ideally adapted to suspensory locomotion. However, the
reorganization of the ulnar side of the carpus.                    hylobatid geometry is better considered as a uniquely
Exploring Phylogenetic and Functional Signals in Complex Morphologies: The Hamate of Extant Anthropoids as a Test-Case Study
214                                                         
                                                         ALMECIJA ET AL

                 Fig. 1. Line drawing depicting a left human hand skeleton in dorsal view. The hamate is highlighted in
               color and its ulnar (upper) and distal (bottom) views are shown. The hamulus, as well as the triquetral,
               fourth and fifth metacarpal (mc) facets are indicated.

derived modification that facilitates midcarpal rotation            ences amongst different locomotor categories in primates
during their highly specialized style of ricochetal bra-            (e.g., Sarmiento, 1988; Spoor et al., 1991). Other studies
chiation (e.g., Jenkins, 1981; Kivell et al., 2013).                have employed multivariate statistical analyses to
  Wrist morphometric analyses, such as individual                   explore morphometric variation in the wrist in compari-
shape ratios based on traditional linear measurements               son to a priori assigned locomotor groups (Begun and
and angles have been used in the past to describe differ-           Kivell, 2011). However, the hamate, like other wrist
Exploring Phylogenetic and Functional Signals in Complex Morphologies: The Hamate of Extant Anthropoids as a Test-Case Study
SHAPE ANALYSIS OF EXTANT ANTHROPOID HAMATES                                         215
bones with complex shapes and multiple articular surfa-          In this study, we use 3D coordinate shape data
ces for surrounding bones, has proved difficult to charac-    derived from 3D surface scans to accomplish four main
terize numerically at the species and even genus level.       objectives:
Alternative approaches based on three-dimensional (3D)           (1) Characterize the complex morphology of the
quantification of carpal and tarsal anatomy have been         hamate among extant anthropoid primates in an effort
particularly useful at discriminating among closely           to distinguish among taxa relying solely on their major
related taxa (Tocheri et al., 2003, 2005, 2007; Marzke        patterns of shape variation. Based on previous studies,
et al., 2010; Tocheri et al., 2011; Dunn et al., 2014). For   we hypothesize that hominoids will show a larger degree
example, this approach was used to compare the                of shape diversity in comparison to platyrrhine and cer-
hamates of fossil hominin taxa within the context of          copithecoid monkeys. In comparison to hominoids, pla-
modern great ape-human variation (Orr et al., 2013).          tyrrhine and cercopithecoid monkeys are expected to
   Since the 1990s, the ongoing development of three-         share larger metacarpal articular surfaces extending
dimensional geometric morphometrics (3DGM) has                onto smaller hamuli facilitating dorsiflexion at the car-
enabled the simultaneous analysis and visualization of        pometacarpal joints.
complex morphology (Bookstein, 1991; Rohlf and Mar-              (2) Test for the presence of phylogenetic structure in
cus, 1993; Dryden and Mardia, 1998; Hammer and                hamate shape and size (i.e., are closely related taxa
Harper, 2006). This method quantifies shape more com-         being more similar to one another than more distantly
prehensively than two-dimensional geometric morpho-           related taxa?). We expect hamate form (i.e., shape plus
metrics (2DGM) and other methods, and may often               size) in hominoids to be different from catarrhine and
capture subtle shape differences between specimens            platyrrhine monkeys, with hylobatids being the most
within a species or among different taxa. Subsequently,       divergent of all apes in terms of shape.
the shape differences can be intuitively visualized as           (3) Explore the predictable covariation of hamate
“warpings” of 3D surfaces generated from thin-plate           shape and size among taxa to assess if shape differences
spline algorithms (Wiley et al., 2005). Several recent        across taxa are related to size differences. Although we
examples exist in anthropology using 3DGM to charac-          expect hylobatids and monkeys to differ in both shape
terize human and primate postcranial elements (e.g.,          and size relative to great apes, it is unlikely that differ-
Turley et al., 2011; Almecija et al., 2013; Bastir et al.,   ences in overall hamate shape will be caused by differen-
2013; Sylvester, 2013; Tallman et al., 2013; Turley and       ces in size alone. To our knowledge, predictability of that
Frost, 2013; Knigge et al., this volume).                     covariation has not been inspected in a comparable data-
   Geometric morphometric analyses characterize com-          set of anthropoid primates (but see Orr et al., 2013 for
plex shapes and reveal overall phenetic affinities effec-     an analysis of specific univariate features and hamate
                                                              size in great apes and humans).
tively. However, the functional significance of shape
                                                                 (4) Test if (and how) hamate shape covaries with a
differences captured by 3DGM is often difficult to inter-
                                                              matrix of known locomotor behaviors to examine
pret. For instance, the observed shape differences often
                                                              whether shape differences (overall) can be reasonably
lack the explicitly biomechanical foundation that typically
                                                              interpreted in a functional context. We hypothesize that
underlies studies based on linear or other 3D measure-
                                                              shape differences between monkey, hylobatid and great
ments taken to represent specific elements of a particular
                                                              ape taxa are related to both phylogenetic divergence and
mechanical model (e.g., load and lever arms, torques, etc.)   associated differences in locomotor behaviors. Hence, we
(Terhune, 2013; Knigge et al., this volume). Nevertheless,    expect a significant portion of total shape variation to
3DGM analyses can use coordinate data that contain bio-       significantly covary with predominant locomotor behav-
mechanically relevant dimensions to help interpret the        iors. For example, given previous observations in the
visual results from a functional point of view (e.g.,         specialized type of locomotion and wrist morphology of
Almecija et al. 2013; Figueirido et al., 2013). A common     hylobatids, we expect a large amount of the total anthro-
procedure is to rely on multivariate statistical analyses     poid hamate shape variance to significantly covary with
(e.g., discriminant function analysis, MANOVA, etc.) to       locomotor behaviors (i.e., due at least to brachiation in
test for differences among “functional” groups defined a      hylobatids).
priori (e.g., Polly, 2008). This is similar to the way that
multivariate analyses are conducted using more tradi-
                                                                      MATERIALS AND METHODS
tional measurements. Another approach is to perform a
multivariate regression or multivariate multiple regres-      Sample and Data Collection
sion of shape coordinate data and, for example, locomotor        We sampled 253 hamates representing 18 extant gen-
categories, taxonomic groups, substrate use, body masses,     era, including platyrrhines, cercopithecoids, hylobatids,
etc. and inspect how dependent variables of interest relate   and hominids (great apes and humans) (Fig. 2; Table 1).
to each other and to the shape coordinates. Degree and        Hominoid taxa (i.e., hylobatids and hominids) were
pattern of covariation can be assessed by comparing           sampled more extensively to capture their patterns of
angles between the projected vectors of different regres-     variation, whereas monkey taxa (especially platyrrhines)
sion analyses (e.g., Turley et al., 2011; Rein and Harvati,   were included as comparisons to help better contextual-
2013; Turley and Frost, 2013). However, results of regres-    ize the hominoid groups. To capture the overall form
sion approaches are still difficult to interpret when dif-    (i.e., shape plus size) of the anthropoid hamate, we col-
ferent locomotor categories (or covariates) are correlated    lected 23 landmarks (Fig. 3; Table 2) from the surfaces
with each other. For example, hypothetical shape changes      of 3D digital hamate models using Landmark Editor
correlated with “terrestriality” may not be clearly           software (ver. 3.6) (Wiley et al., 2005). The 3D models
separated from those correlated with general                  were obtained using laser scanning (Nextengine, Scan-
“quadrupedalism” (whether it is arboreal or terrestrial).     studio HD Pro), with the exception of three small
Exploring Phylogenetic and Functional Signals in Complex Morphologies: The Hamate of Extant Anthropoids as a Test-Case Study
216                                                          
                                                          ALMECIJA ET AL

                 Fig. 2. Time-calibrated phylogenetic tree showing the relationships among the modern anthropoid pri-
                mates sampled in this study. Representative hamates are presented in dorsoulnar view for selected taxa.

platyrrhine hamates (Aotus, Saguinus, Saimiri) that                  nents of the covariance matrix of the original variables,
were acquired using micro-CT imaging with surfaces                   such that the new shape space was centered on the aver-
extracted using the segmentation tool of Avizo software              age shape configuration. Thus, the first principal compo-
(ver. 7.0). All raw scans were edited with Geomagic                  nent axis runs through the major axis of variation in the
(vers. 12–2013) to clean the meshes, fill holes, and                 data with subsequent axes running at orthogonal angles
standardize all hamate models to a similar number of                 through the minor axes of variation, implying a rigid
vertices. It has been shown that scale-free data collected           rotation of the original variables (Polly, 2008; Polly
from 3D surface models obtained from different sources               et al., 2013). Shape differences among our extant sample
are reasonably compatible (Tocheri et al., 2011). All final          of anthropoid primates were tested by means of multi-
3D models were opened in Landmark Editor for data                    variate analysis of variance (MANOVA) of the first three
collection.                                                          principal components (which concentrate 53.4% of the
   The phylogenetic relationships of the taxa examined               variance). Post hoc significant differences among platyr-
(Fig. 2) were recovered from a chronometric consensus                rhines, cercopithecoids, hylobatids, the three great ape
tree based on molecular data and downloaded from “The                genera and modern humans was assessed through
10KTrees Project” (ver. 3; http://www.10ktrees.fas.har-              Hotelling’s P values (Bonferroni’s corrected).
vard.edu) (Arnold et al., 2010). Some nomenclature, such                To assess the phylogenetic patterning in the shape
as the new generic name (Hoolock) for hoolock gibbons                space, we mapped a phylogenetic tree onto the shape
(Mootnick and Groves, 2005), was updated to reflect                  space defined by the two first axes of a PCA of the covar-
recent taxonomic revisions.                                          iance matrix of extant taxa means. Hypothetical ances-
                                                                     tral states (internal nodes in the tree) were
                                                                     reconstructed using squared-change parsimony (Maddi-
Analyses
                                                                     son, 1991) with a method developed for geometric data
   Shape data were obtained from raw coordinates                     in which shape is treated as a single multidimensional
through a full (generalized) Procrustes fit analysis—                character and weighed by branch length (Klingenberg
which rotates, translates and size-scales the landmark               and Gidaszewski, 2010). Subsequently, the estimated
configurations to unit of centroid size—and posterior                ancestral node configurations were plotted on the origi-
orthogonal projection on to the tangent space (Dryden                nal shape space, and the branches of the tree were con-
and Mardia, 1998). Major patterns of hamate shape vari-              nected to obtain a phylomorphospace (Rohlf, 2002;
ation were explored through a principal components                   Sidlauskas, 2008; Monteiro, 2013). The previous steps
analysis (PCA). The original coordinate system of the                were necessary to subsequently test for the presence of a
shape space was transformed to the principal compo-                  phylogenetic signal in hamate shape space by means of
Exploring Phylogenetic and Functional Signals in Complex Morphologies: The Hamate of Extant Anthropoids as a Test-Case Study
SHAPE ANALYSIS OF EXTANT ANTHROPOID HAMATES                                      217
                               TABLE 1. Sample of hamates analyzed in this study
Genus                        Species/subspecies              Female           Male           Unknown             Total
Homo                    Homo sapiens                            8              33                14              55
Pan                     Pan paniscus                           10              11                1               22
                        P. troglodytes troyglodytes             5              5                 1                11
                        P. troglodytes schweinfurthii           0              2                 0                2
                        P. troglodytes verus                    1              2                 0                3
                        P. troglodytes ssp.                     2              2                 1                5
Gorilla                 Gorilla gorilla                         7              15                1               23
                        G. beringei beringei                    4              4                 3                11
                        G. beringei graueri                     6              11                3               20
Pongo                   Pongo pygmaeus                          7              7                 0               14
                        Pongo abelii                            1              4                 0                5
                        Pongo sp.                               0              1                 0                1
Symphalangus            Symphalangus syndactylus                4              6                 0               10
Hoolock                 Hoolock hoolock                         2              3                 0                5
Hylobates               Hylobates klossi                        5              0                 0                5
                        Hylobates lar                           2              1                 0                3
Nomascus                Nomascus leucogenys                     0              1                 0                1
Papio                   Papio anubis                            4              5                 0                9
                        Papio cynocephalus                      1              1                 0                2
Macaca                  Macaca mulatta                          5              5                 0               10
Nasalis                 Nasalis larvatus                        3              5                 0                8
Alouatta                Alouatta seniculus                      4              4                 4               12
Ateles                  Ateles belzebuth                        1              0                 0                1
                        Ateles fuscieps                         0              2                 0                2
                        Ateles geoffroyi                        2              0                 1                3
                        Ateles paniscus                         1              2                 0                3
                        Ateles sp.                              1              0                 0                1
Brachyteles             Brachyteles arachnoides                 0              0                 1                1
Lagothrix               Lagothrix lagothrica                    0              0                 1                1
                        Lagothrix sp.                           0              0                 1                1
Aotus                   Aotus trivirgatus                       0              0                 1                1
Saguinus                Saguinus oedipus                        0              0                 1                1
Saimiri                 Saimiri sciureus                        0              0                 1                1
                        Total sample                                                                             253

a permutation test approach (Laurin, 2004) extended for     shape and locomotor categories in selected taxa (Rohlf
multivariate analysis of coordinate data (Klingenberg       and Corti, 2000). This method differs from multivariate
and Gidaszewski, 2010). This technique simulates the        regression in that both the shape and covariate blocks
null hypothesis of complete absence of phylogenetic         are treated symmetrically rather than as one set of vari-
structure among hamate shapes in extant anthropoids.        ables (X axis) being used to predict variation in the
The species mean shape configurations were randomly         other set of variables (Y axis). These new pairs of varia-
distributed as the TIPS of the phylogeny in 10,000 per-     bles (the different PLSes) represent linear combinations
mutations. For each permutation, tree length (i.e., the     of variables within the original two sets (the blocks).
sum of the squared Procrustes distances between ances-      These linear combinations are constructed so that the
tral and descendant shapes for all branches) was com-       new variables account for as much of the covariation as
puted. If more than 5% of the resulting tree lengths        possible between the two original sets with the goal to
computed in the permutation test were greater than the      find relationships between them without assuming that
one obtained with the original data, the null hypothesis    one is the cause of the variation in the other. In our
of absence of phylogenetic structure in the data was        case, one block was constituted by hamate shape (Pro-
rejected.                                                   crustes coordinates) and the other by a matrix of the
   Covariation of hamate shape and size (i.e., allometry)   frequencies of use for several types of arboreal locomo-
was explored using multivariate regression of all the       tion by extant hominoid and selected anthropoid species
Procrustes coordinates on log-transformed centroid size     (Table 3). Data on arboreal locomotion was used because
(logCS), using both species mean configurations (i.e.,      it is available for more taxa (Thorpe and Crompton,
TIPS of the phylogeny) and phylogenetic independent         2006; their Table 7 and references therein). The RV
contrasts (PICs; Felsenstein, 1985). To assess the reli-    coefficient, a multivariate analogue of the squared corre-
ability of these multivariate regressions, a permutation    lation (Escoufier, 1973), was used as an overall measure
test (10,000 iterations) against the null hypothesis of     of association between the two blocks, and a permuta-
independence between the dependent (hamate shape)           tion test (10,000 iterations) was employed against the
and independent (logCS) variables was performed.            null hypothesis of complete independence. In addition, a
   We also performed a two-block partial least squares      second PLS analysis was carried out on the PICs for
(PLS) analysis to test for covariation between hamate       both blocks of variables to test if covariation between
Exploring Phylogenetic and Functional Signals in Complex Morphologies: The Hamate of Extant Anthropoids as a Test-Case Study
218                                                       
                                                       ALMECIJA ET AL

               Fig. 3. Left hamate of Pan troglodytes illustrating the 23 surface landmarks employed in this study.
              Warped versions of the same three-dimensional hamate model are used in subsequent figures.

hamate shape and arboreal locomotion was detected                    To visualize morphological variation associated with
when considering phylogenetic structure. The amount of            each axis in each analysis, coordinates of shape change
shape variance explained by each PLS axis was com-                along the axes were imported into Landmark Editor and
puted by dividing the “singular value” (or “singular              used to warp (using thin-plate splines) a 3D hamate
warp”) of each PLS by total variance within the shape             model of Pan troglodytes (shown in Fig. 3) into desired
block (sum of eigenvalues, computed after a PCA of the            score values. All the statistical analyses were carried out
shape data).                                                      using MorphoJ (ver. 1.05; Klingenberg, 2011), PAST
SHAPE ANALYSIS OF EXTANT ANTHROPOID HAMATES                                        219
     TABLE 2. Landmarks employed in this study to           Our analyses also show significant differences between
         capture hamate external morphology                 hamate shape in platyrrhines and cercopithecoids
#        Type                   Description                 (P < 0.05).
                                                               PC1 (25.8% of total variance) is related to overall pro-
1        II       Junction of triquetral and capitate       portions of the hamate body and hamulus, as well as to
                    surfaces on the proximopalmar side
                                                            the relative depth of the fourth and fifth metacarpal
2        II       Point of maximum inflexion on the
                    palmar border of capitate surface       articulations. This axis generates four relatively distinct
3        II       Junction on the palmar border of the      clusters of hylobatids, monkeys, Pan/Pongo, and
                    capitate and metacarpal IV surfaces     Gorilla/Homo from left (negative) to right (positive)
4        II       Junction of the triquetral/lunate and     (Fig. 4B,C). Most Ateles specimens are near the negative
                    capitate surfaces on the dorsal side    extreme of the monkey cluster, overlapping with hyloba-
5        III      Midpoint between 4 and 6 on the dorsal    tids. Hylobatids have hamate bodies and hamuli that
                    border of the capitate surface          are proximodistally long, and radioulnarly and dorsopal-
6        II       Junction of the dorsal border of the
                    capitate and metacarpal IV surfaces     marly narrow (Fig. 4D). The fifth metacarpal facet proj-
7        II       Proximal root of the hamulus              ects distally over that of the fourth metacarpal while
                    on the palmar side                      the medial side of the hamate is oriented obliquely. Con-
8        II       Most palmarly protruding point            versely, human and gorilla hamates are proximodistally
                    of the hamulus                          short, and radioulnarly and dorsopalmarly wide, with a
9        II       Most distally protruding point            sigmoid medial side and a more vertical overall orienta-
                    of the hamulus                          tion of the bone.
10       II       Junction of metacarpals IV and V
                    on the palmar side                         PC2 (15.3% of variance) is also related to the radioul-
11       II       On the triquetral surface,                nar width of the hamate body, along with the size and
                    distodorsal border                      orientation of the hamulus and the articular surfaces
12       III      Midpoint between 11 and 13                for the triquetrum and metacarpals. Along this axis,
13       II       On the triquetral surface,                hylobatids are maximally separated from monkeys, with
                    distopalmar border                      hominid taxa falling in between these two extremes,
14       II       On the triquetral surface, most           although humans overlap more with monkeys than do
                    convex point of the dorsal side
15       II       Midpoint of the triquetral surface        the other hominids (Fig. 4B). Negative values corre-
16       II       On the triquetral surface,                spond to radioulnarly narrow hamates with large
                    point of maximum inflexion              hamuli that project distally, triquetral facets that are
                    on the palmar side                      dorsopalmarly wide proximally and dorsopalmarly nar-
17       II       Most proximal point on the junction       row distally, and fifth metacarpal facets that are radio-
                    of triquetral and capitate surfaces     ulnarly narrow relative to those of the fourth
18       II       Radiopalmar border on the                 metacarpal. In contrast, positive values on this axis
                    metacarpal IV surface
19       II       Ulnopalmar border on the
                                                            describe hamates with a radioulnarly wide body, a small
                    metacarpal V surface                    and less distally projecting hamulus, a triquetral facet
20       III      Midpoint between 10 and 22,               that is dorsopalmarly narrow proximally and distally
                    following the border between            but widest at its midpoint, and a fifth metacarpal facet
                    the surfaces for metacarpals IV and V   that that is larger than that of the fourth metacarpal
21       III      Midpoint between 19 and 23                and extends onto the hamulus. Together, PC1 and PC2
22       II       Junction of metacarpals IV and V          separate hominids from both hylobatids and anthropoid
                    on the dorsal side
                                                            monkeys (Fig. 4B).
23       II       Ulnodorsal border on the
                    metacarpal V surface                       PC3 (12.3% of variance) is related to hamate body
                                                            proportions, the shape and distal projection of the
The description of the landmark type is after Bookstein     hamulus, and the outline shape of the metacarpal
(1991).                                                     articulations. This axis clearly separates gorillas from
                                                            modern humans (negative and positive values respec-
                                                            tively), while also acting to separate Pan and monkeys
(ver. 3.01; Hammer et al., 2001) and SPSS software          (negative values) from Pongo and hylobatids (positive
packages (ver. 17).                                         values). Negative values correspond to radioulnarly
                                                            wider and proximodistally shorter hamates (but to a
                 RESULTS                                    smaller extent than seen in PC 1 and 2), distally pro-
                                                            jecting hamuli and a trapezoidal outline for the meta-
Hamate Shape Variation Among Extant
                                                            carpal facets with its larger dimension located dorsally.
Anthropoids                                                 In addition, the fifth metacarpal facet extends distally
  Overall, humans, gorillas, hylobatids, and monkeys        onto the hamulus. Positive values correspond with
(and to a lesser degree chimps and orangutans) can          hamate bodies that are slightly longer proximodistaly
be differentiated from one another based on their           and slightly narrower radioulnarly, hamuli that extend
hamate shapes when the three first PC axes (account-        palmarly and a square-shaped articular outline for the
ing for 53.4% of the variance, Fig. 4D) are inspected       metacarpals. The fifth metacarpal facet does not
together (Fig. 4A). Post hoc comparisons reveal that        extend on to the hamulus. Overall, the combination of
monkeys, hylobatids, Pongo, Gorilla, Pan, and H. sapi-      PC1 and PC3 (Fig. 4C) separates humans, hylobatids
ens are statistically different from each other based on    and gorillas (with the exception of two specimens)
the variation along the three first axes (P < 0.0001).      from the remaining taxa.
220                                                    
                                                    ALMECIJA ET AL

      TABLE 3. Proportion of arboreal locomotion used in the two-block partial least squares (PLS) analysis
Taxon                                   QTW             VCD            BW             OCT             BFS             DL
                                                                                           a
Pan paniscus                             0.32           0.53           0.01          0.255            0.09           0.04
Pan troglodytes schweinfurthii           0.36           0.49           0.07          0.05             0.05           0.00
Pan troglodytes verus                    0.22           0.68           0.03          0.08a            0.07           0.01
Gorilla beringei beringei                0.53           0.4            0.02          0.06a            0.05           0.00
Gorilla gorilla gorilla                  0.19           0.48           0.05          0.17             0.03           0.01b
Pongo pygmaeusc                          0.18           0.26           0.07          0.22             0.13           0.01
Symphalangus syndactylus                 0.00           0.32           0.08          0.00             0.59           0.02
Hylobates lard                           0.01           0.16           0.02          0.00             0.67           0.14
Papio Anubis                             0.68           0.21           0.00          0.00             0.00           0.10
Ateles belzebuth                         0.21           0.13           0.01          0.28             0.22           0.02
Lagothrix lagotricha                     0.29           0.14           0.00          0.30             0.09           0.04

QTW, quadrupedal and tripedal walk; VCD, vertical climb and descent; BW, bipedal walk; OCT, orthograde clamber and
transfer; BFS, brachiation and forelimb swing; DL, drop and leap.
Data were obtained from Table 7 in Thorpe and Crompton (2006). Their percentages are presented here as a proportion.
a
  Average data for males and females, obtained from Carlson (2005).
b
  Unknown, but because drop and leap probably constitute a very small component of western gorillas locomotor repertoire,
it was approximated at 1%.
c
  Thorpe and Crompton (2006) study, omitting juveniles.
d
  Data reported in Thorpe and Crompton (2006) for gibbons corresponds to pooled data for H. agilis, H. lar and H. pileatus
from different sources. Because data come from different sources, and not all arboreal behaviors reported in the original
studies were included, percentages often do not add to 100%.

Phylogenetic Structure in the Extant                           suggests that the relationship is mostly artifactual,
Anthropoid Hamate Shape and Size                               being caused by the presence of three platyrrhine spe-
                                                               cies that are comparatively small (Saimiri, Saguinus,
   Shape space patterns based on species means reason-         Aotus). This is confirmed by the regression of the phylo-
ably display a phylogenetic structure (Fig. 5). This is        genetically independent contrasts (PICs) of hamate
confirmed by the results of our permutation test               shape and logCS, revealing no significant relationship
(P < 0.0001), allowing us to reject the null hypothesis of     (P 5 0.8; Fig. 7B). This result indicates that differences
complete lack of phylogenetic signal. However, the maxi-       in hamate shape among extant hominoids and other
mum axis of variance in shape space (i.e., PC1) is             anthropoids are not merely “explained” or predicted by
defined by monkeys (especially some platyrrhine species)       their differences in size.
and hylobatids, respectively, with great apes in an inter-
mediate position. Branch lengths indicate that, not only
within hominoids but also within the whole sample of           Locomotor Signatures in the Hamate
anthropoid species examined here, most of the shape               The results of PLS show that there is a strong and
changes occurred in the branch (Fig. 5, marked in red)         significant overall association between hamate shape
connecting the hypothetical ancestor of all hominoids          and the types of arboreal locomotion (RV 5 0.73; P <
(LCAH) and that of hylobatids. Modern hominids exhibit         0.0001). The relationship stands (although not as
fewer shape differences in their hamates than do hyloba-       strongly) when considering phylogenetic structure
tids (i.e., they have changed less) since their divergence     through the use of PICs (RV 5 0.57; P < 0.05). Figure 8
from the LCAH. This phylomorphospace (reconstructed            shows the results for the first two PLS that account for
based solely on the extant taxa included) suggests that        98% of the total covariance.
since the LCAH, the hamates of modern humans and                  Along PLS1 (79.5% of total covariation and 99% of the
gorillas have “evolved more” than that of chimps (i.e.,        total shape variation), negative values in the arboreal
common chimpanzees and bonobos), which are recon-              locomotion block are related to brachiation and arm-
structed by this analysis as having secondarily evolved        swinging, and covary most with proximodistally long
into a morphology similar to that of orangutans (thus          and radioulnarly narrow hamate bodies, as well as with
indicating homoplasy). Orangutan hamate shape                  large and distally extending hamuli and convex trique-
appears more conservative among great apes. Similarly          tral facets. Conversely, positive values in the locomotor
as with hamate shape, when hamate centroid size is             block are mostly related to “quadrupedal/tripedal walk”
mapped onto our anthropoid phylogeny (Fig. 6), differen-       behaviors and to a lesser degree to “vertical climbing
ces in size are evident among taxa and phylogenetic            and descent” modes of locomotion. These latter arboreal
signal is further detected (P < 0.0001).                       locomotor types covary most in the hamate shape block
                                                               with proximodistally shorter and radioulnarly broader
Predicable Covariation of Hamate Shape and                     hamates, with less distally protruding hamuli and
                                                               “spiral” shaped triquetral facets.
Size
                                                                  Along PLS2 (18.2% of total covariation and 47.3% of
   The multivariate regression of shape and log-               the total shape variation), negative values covary most
transformed centroid size (logCS) based on species             with “orthograde clamber and transfer” behaviors and to
means (TIPS) shows a significant relationship (P < 0.05),      a lesser extent with “quadrupedal/tripedal walk”,
although accounting only for minimal portion (8.4%) of         whereas in the hamate shape block they relate to proxi-
total shape variation (Fig. 7A). An inspection of the plot     modistally short and radioulnarly wide hamates, with
Fig. 4. Hamate shape variation among extant anthropoid primates.        iance accounted for each PC and warped surfaces (based on a thin-
A: Major taxonomic groups are differentiated along the first three prin-   plate spline) representing shape changes associated with each axis (in
cipal components (PCs). B: Plot showing PC1 vs. PC2 (25.8% and             ulnar, dorsal, and distal views, respectively). PC1 represents values of
15.3% of variance explained, respectively). C: Plot showing PC1 vs.        20.20 and 10.20; PC2 represents values of 20.20 and 10.20; PC3
PC3 (12.3% of variance explained). D: Percentages of the total var-        represents values of 20.15 and 10.15.
222                                                            
                                                            ALMECIJA ET AL

  Fig. 5. Phylomorphospace of the anthropoid hamate. A: The phylog-    extant species average configurations (in dorsoulnar view) were plot-
eny presented in Fig. 1 is projected onto the shape space defined by   ted. Phylogenetic structure is detected in this shape space (tree
the two first principal components (PCs) of an analysis performed on   length: 0.22; P < 0.0001). LCAH indicates the reconstructed position
the covariance matrix amongst extant species means. Internal node      for the last common ancestor of all modern hominoids. The longest
morphologies were reconstructed using squared-change parsimony         branch (red marked) is leading from the LCAH to the hylobatids. B:
and branch lengths were derived from estimated divergence times. To    Percentages of the total variance accounted for each PC.
facilitate the morphospace interpretation, 3D surfaces representing

large facets for the metacarpals (especially the fifth),               acterized by means of their major trends of hamate
small and nonprojecting hamuli and dorsopalmarly short                 shape variation (Fig. 4A). By relying on the two major
triquetral facets on the proximal portion. In contrast,                axes of variation (Fig. 4B), extant hominoid hamates are
positive values that mostly covary in the arboreal loco-               readily distinguishable from those of extant monkeys,
motor block are related to “vertical climb and descent”                and in turn, those of hylobatids can be distinguished
behavior. In the hamate shape block, this corresponds to               from those of great apes and humans. Within hominidae,
slightly proximodistally longer and radioulnarly nar-                  human and gorilla hamates are more easily differenti-
rower hamates, with larger and distally projecting                     ated compared with those of chimps and orangutans
hamuli, and larger and more globular proximal portions                 (Fig. 4A,C). This observation might be explained by sym-
of the triquetral facets (thus forming a true spiral facet).           plesiomorphy of the latter two in comparison to the
   When the distribution of taxa in the PLS bivariate                  more derived morphologies of humans and gorillas (Fig.
plots is inspected (Fig. 8), overall, closely related taxa             5) as similarly observed by others (Kivell et al., 2013).
occupy a similar position in the scatter, similarly to the
anthropoid hamate shape space (Figs. 4, 5).                            Inspection of the Phylogenetic Structure in
                                                                       Anthropoid Hamate Shape
                 DISCUSSION                                               As explained above, major shape differences exist
Characterization of Hamate Morphology Among                            among major taxonomic groups, and thus a significant
                                                                       phylogenetic pattern is detected (Fig. 5). Not only is the
Extant Anthropoids
                                                                       colobine Nasalis similar to the papionin taxa, in accord
   Our 3DGM results largely agree with previous obser-                 with prior work (Lewis, 1972; O’Connor, 1975), but pla-
vations on the relative uniformity of hamate shape                     tyrrhines also overlap almost completely in shape space
within cercopithecoids, as well as the extensive overlap               with cercopithecoids despite their early divergence time
with platyrrhine taxa (Fig. 4). At the same time, when                 (Fig. 2). In contrast, extant hominoids—which are more
compared at the species level, platyrrhines exhibit con-               closely related to each other than to platyrrhines or cer-
sistent shape differences vis-
                              a-vis cercopithecoids (Fig.              copithecoids—occupy a larger portion of the shape space
5), although more species and specimens per taxa are                   (i.e., higher shape disparity). This is remarkable consid-
necessary to formally confirm this observation. However,               ering that this analysis incorporates fewer terminal taxa
overall, major anthropoid taxonomic groups can be char-                from the hominoids than from monkeys (12 vs. 14
SHAPE ANALYSIS OF EXTANT ANTHROPOID HAMATES                                           223

                  Fig. 6. Anthropoid phylogeny mapped on hamate centroid size. A significant phylogenetic signal is
                detected (tree length: 848.88; P < 0.0001).

respectively). These results suggest that hominoids, as a          and some type of climbing, but not below-branch suspen-
group, have experienced increased selective pressures on           sion (e.g., Ward et al., 1993; Madar et al., 2002; Ishida
wrist morphology in comparison to other anthropoids.               et al., 2004; Moy  a-Sola et al., 2004; Ward, 2007),
The hominoid’s larger hamate shape disparity is in large           although some workers have inferred knuckle-walking
part due to the highly distinctive hylobatids; the longest         for Sivapithecus (Begun and Kivell, 2011).
branch in the hominoid part of the tree occurs between                The shape space defined by anthropoid primate
the last common ancestor of all hominoids (LCAH) and               means (Fig. 5) is similar to that of individual speci-
that of all hylobatids (Fig. 5, marked in red), suggesting         mens (Fig. 4B) although it shows an inversion of the
a large amount of evolutionary change in the lineage               two major axes of variance (i.e., PC1 in the first anal-
leading to gibbons and siamangs. Similar results have              ysis is PC2 in the second and vice versa). There is a
been found recently when relying on a Mosimann shape               shape shift in modern hominoids relative to monkeys.
space and a phylogenetic PCA (Kivell et al., 2013). These          This shift involves an enlargement of the hamulus
results support the hypothesis that hylobatids are highly          with a concomitant reduction of the relative size of
autapomorphic in their hamate morphology (more so                  the metacarpal facets, as well as a dorsopalmar
than great apes relative to monkeys). Derived wrist mor-           enlargement of the proximal portion of the triquetral
phology in hylobatids is probably related to their very            facet, which also happens to be more globular (see
specialized ricochetal brachiation (e.g., Jenkins, 1981),          changes in PC2 of Fig. 4). This latter morphology
which could be considered part of an early hominoid                may be related to the “concavo-convex spiral complex”
radiation in a new adaptive zone.                                  described in extant great apes by Lewis (1972; see
   Fossil evidence also strengthens the view that hyloba-          discussion below). As for the enlargement of the
tids are too derived in their wrist joints to be considered        hamulus in hominoids, the small (and likely plesio-
good ancestral models for hominoids. Based on the study            morphic) hamulus described for Proconsul (Beard
of the wrist bones of Proconsul, Lewis (1972) observed             et al., 1986) and Sivapithecus (Spoor et al., 1991) sug-
morphological traits enhanced for greater mobility in              gests that its size increased independently in hyloba-
comparison to monkeys and hypothesized that an early               tids and extant great apes. The results of our PLS
feature defining hominoids in their new ecological niche           (Fig. 8) reinforce the idea that the small hamuli of
different from other catarrhine primates was related to            fossil apes are related to quadrupedalism (PLS1) and/
“brachiation”. However, he admitted that the term bra-             or even orthograde clambering (PLS2).
chiation, with the connotations of a highly derived                   Extant great apes (especially gorillas) and humans
gibbon-like behavior, was probably not the most                    occupy a region of the hamate shape space that is char-
adequate term to describe the entire repertoire of arbo-           acterized by dorsopalmarly extended hamates (and
real activities displayed by early hominoids. In fact, lat-        hamuli) that are also less obliquely oriented. This is also
ter studies have found that, on the basis of their total           evidenced by the more nearly coplanar positioning of the
morphological patterns, early and middle Miocene apes              metacarpal facets (i.e., the fifth metacarpal facet does
relied heavily on arboreal palmigrade quadrupedalism               not extends as far distally beyond the fourth when the
Fig. 7. Predictable covariance of anthropoid hamate shape and size. A: Anthropoid species average
shapes regressed on log-transformed centroid size (TIPS) reveals a significant relationship (P < 0.05)
although only accounting for 8.4% of the total shape variation. B: Regression of phylogenetic independ-
ent contrasts (PICs) of shape and log-transformed centroid size reveals no significant relationship
(P 5 0.8).
SHAPE ANALYSIS OF EXTANT ANTHROPOID HAMATES                                                         225

   Fig. 8. Two-block partial least squares (PLS) analysis of hamate     axis as well as a biplot with the PLS coefficients for the arboreal loco-
shape and arboreal locomotor categories revealing a significant asso-   motion block. D: PLS coefficients for the hamate shape block are dis-
ciation (RV 5 0.73; P < 0.0001). A: PLS1 scores of arboreal locomo-     played as shape changes corresponding to PLS scores of (20.2 and
tion vs. hamate shape. B: PLS2 scores of arboreal locomotion vs.        10.1) in each axis.
hamate shape. C: Total amount of covariance contained in each PLS

radial side of the bone is aligned vertically as it does in             becomes statistically nonsignificant when studying their
monkeys and hylobatids; see PC1 in Fig. 4).                             phylogenetic independent contrasts (Fig. 7B). This indi-
   Since the reconstructed LCAH, orangutans (followed                   cates that the observed shape differences between big
by chimps) seem to possess the most plesiomorphic                       taxonomic groups are not simply correlated with overall
hamate shape, while gorillas and humans with stockier                   size. However, it could be argued that some of the loco-
hamate bodies and palmarly extending hamuli display                     motor behaviors displayed by different anthropoid taxa
the most derived hamates within extant Hominidae, in                    analyzed here are more or less compatible with a certain
agreement with recent findings using similar methods                    body mass. For example, among other factors, brachia-
(Kivell et al., 2013). While the hamate morphology of                   tion is facilitated in hylobatids because they are small
gorillas can be intuitively associated with weight bearing              (e.g., Preuschoft and Demes, 1985; Swartz, 1989). In
in these large-bodied terrestrial hominoids (Sarmiento,                 that sense, certain behaviors could show some amount of
1994), the morphology of the human hamate (freed from                   significant covariation with size (predictable or not) and
selective locomotor pressures) should be explored under                 in turn with specific aspects of shape (see following sec-
the scope of enhanced manipulation. Future analyses                     tion). Future analyses (Almecija et al., in preparation)
including fossil species near major evolutionary splitting              will inspect if and how patterns of hamate shape varia-
events (e.g., Almecija et al., 2013) and employing phylo-              tion within hominoid species are predictably correlated
genetically weighted PCA (e.g., Kivell et al., 2013; Polly              with size.
et al., 2013) may allow a better determination of the
polarity and degree of shape change within each lineage,
and thus properly allow us to assess and resolve the pat-               Exploring Covariation of Hamate Shape and
terns and tempo of hominoid hamate evolution.                           Locomotion
                                                                           Most of the hamate functional morphology discussion
Is There Predictable Covariation between                                has focused on the shape of the triquetral articular sur-
                                                                        face and the length and orientation of the hamulus. In
Hamate Shape and Size?
                                                                        this section, we discuss how previous observations on
   Although significant phylogenetic patterning exists in               the anthropoid hamate morphology relate to our results.
both hamate shape and size (Figs. 5, 6), a very weak cor-                  Lewis (1972) considered African apes to share a com-
relation of hamate shape and size exists when species                   mon hamate morphology: a narrower version of the
means are compared to each other (Fig. 7A), and even                    wedge-shaped monkey hamate that is oriented with the
226                                                     
                                                     ALMECIJA ET AL

long axis of the body directed proximodistally instead of      terms of the “spiral” complex on the triquetral facet—
obliquely as in monkeys and hylobatids. This morphology        assuming that its morphology is related to the long axis
reduces the articulation with the lunate (increasing artic-    of the hamate body being more proximodistally aligned,
ulation with the capitate) and causes the triquetral facet     with a globular (more convex) proximal end and a more
to be aligned more obliquely. This arrangement causes          concave distal portion—our PC1 of the entire anthropoid
the triquetral facet to be more ulnarly oriented overall,      sample (Fig. 4) captures a gradient (negative to positive)
but distally “inflected” toward the hamulus, resulting in      from hylobatids to monkeys, followed by Pongo/Pan and
the so-called “spirally concavo-convex” facet. Lewis (1972)    finally Gorilla/Homo. Therefore, our results agree with
contended that the “spiral” character of the triquetral        those of Lewis (1972, 1989) in indicating better develop-
facet contributed to the “screw mechanism” of the midcar-      ment of this complex in extant great apes (and humans)
pal complex, working as a “close-packing” system limiting      than in monkeys. However, hylobatids show the most
extension, which he argued is an adaptation to resisting       disparate (autapomorphic) morphology, with no trace of
the tensile forces associated with regular suspension.         “spiralization”. Instead, they exhibit an oblique proximo-
Such screw-clamp kinematics of the carpus has been             distal axis of the hamate with a large and very globular
found in humans and chimpanzees (MacConaill, 1941;             triquetral facet that lacks the distal concavity. This
Orr et al., 2010), but the exact role of triquetrohamate       shape is most likely related to the “ball-and-socket” mid-
form in that mechanism is not clear (Orr, 2010). Further-      carpal morphology that has been widely interpreted as
more, Jenkins and Fleagle (1975) argued that the pres-         an adaptation for their very specialized ricochetal bra-
ence of the spiral concavo-convex complex is common in         chiation (e.g., Tuttle, 1969; Lewis, 1972; Usherwood
quadrupeds such as Macaca, thus falsifying the connec-         et al., 2003). Hylobatids also exhibit relatively proximo-
tion between this trait and below-branch suspension. Sar-      distal long hamates, which Kivell et al. (2013) hypothe-
miento (1994) pointed out that the spiral complex              size is related to their overall long forelimbs (Drapeau
effectively increases the radius of curvature of the carpal    and Ward, 2007) in relation to enhanced suspension.
row proximally and increases the surface area of the dis-      Hylobatids are also the smallest of the modern homi-
tal triquetrohamate joint that is directed orthogonal          noids (Smith and Jungers, 1997), which is probably
transarticular loads when the midcarpus is extended. As        tightly related to their unique capability among homi-
such, he suggests that this character is related to quadru-    noids to perform a true “ricochetal brachiation” involving
pedalism rather than suspension.                               a period of free flight (e.g., Conroy and Fleagle, 1972;
   It has been hypothesized that the overall degree of con-    Preuschoft and Demes, 1985; Swartz, 1989; Hunt, 1991).
vexity of the hamate triquetral facet reflects the potential   Hylobatids therefore define the shape that is related to
range of flexion-extension at the midcarpus (Spoor et al.,     brachiation along the first axis of the PLS analysis (Fig.
1991). This facet is more curved (or globular) in apes and     8A,C).
flatter proximally in terrestrial monkeys, with the latter        Although there is some overlap, Pan and Gorilla dis-
ostensibly preventing extension during terrestrial locomo-     play a more globular proximal triquetral facet (i.e., con-
tion in nonprimate mammals (Yalden, 1971). However,            cavoconvex) than does Pongo (PC3 in Fig. 4C). This may
cadaver-based experiments indicate that although Papio         reflect a wider weight-bearing midcarpal surface in the
has limited mobility at the triquetrohamate joint during       African apes. The complex morphology of the spiral con-
dorsiflexion, palmigrade-capable monkeys exhibit much          cavoconvex morphology found in some monkeys (Jenkins
larger ranges of motion at this articulation than do chim-     and Fleagle, 1975), Proconsul (Lewis, 1972), orangutans,
panzees or orangutans (Orr, 2010).                             and especially some African ape taxa might be explained
   Sarmiento (1988) provided the first quantitative study      in terms of the functional demands of arboreal locomo-
relating the orientation of the triquetral facet (comput-      tion in these forms. These locomotor behaviors require
ing angles between the capitate, triquetral and metacar-       both enhanced mobility (i.e., rotation and ulnar devia-
pal facets) with the capacity of weight transfer through       tion) and stability. The close-packing mechanism pro-
the triquetrohamate joint. As with Lewis, Sarmiento            vided by the spiral facet has been hypothesized to be
related the more ulnarly oriented triquetral facets of         advantageous during knuckle-walking (e.g., Richmond
hominoids such as Pan and Pongo to less effective              et al., 2001), but the distribution of this feature may
weight transmission and wider range of movement at             instead indicate its importance during use of the hand
the midcarpal joint (in comparison to that of monkeys).        in arboreal settings. High positive loadings of vertical
However, he noted that the triquetral facet in gorillas is     climbing associated with a highly convex proximal tri-
nearly parallel to the articular surfaces for the metacar-     quetral facet in PLS2 further suggest that this is the
pals, thus favoring weight support (Sarmiento, 1988,           case (Fig. 8).
1994). Other studies using traditional goniometric meas-          In reference to the greater development of the hamu-
urements (Richmond, 2006) and least-squares planes fit         lus in extant hominoids in comparison to monkeys, it
to segmented joint surfaces of 3D polygon models of the        has been related to the action of the flexor carpi ulnaris
hamate (Orr et al., 2013) had similar results with goril-      (which primarly inserts on the pisiform) but acts on the
las displaying the most proximally oriented triquetral         hamulus by means of the pisohamate ligament (Sch€on
surfaces.                                                      and Ziemer, 1973; O’Connor, 1975; Lewis, 1977). A pal-
   Our PLS results based on arboreal locomotor behav-          marly protruding hamulus favors powerful flexion of a
iors and with PLS1 accounting for 79.5% of shape-              fully extended wrist (Sarmiento, 1988). Conversely, the
arboreal locomotion covariation (and up to 99% of the          hamulus is more distally oriented in hominoid taxa that
total hamate shape variation) also indicate that quadru-       regularly use arboreal supports. A distally oriented
pedal primates exhibit wider hamate bodies with less           hamulus improves the lever arm of the flexor carpi ulna-
globular and more proximally oriented triquetral facets        ris for ulnar deviation, flexion of an already flexed wrist,
and smaller hamuli than do suspensory taxa (Fig. 8). In        or preventing dorsiflexion (Sarmiento, 1988, 1994; Patel
SHAPE ANALYSIS OF EXTANT ANTHROPOID HAMATES                                        227
et al., 2012); thus, it is probably related to increasing      253 anthropoid primates using principal components
wrist torque potential during climbing or suspension.          analysis on the covariance matrix of 23 Procrustes-
Sarmiento (1994) pointed out that the hamulus of goril-        aligned surface landmarks. This approach allowed us to
las is better developed than in chimpanzees, with a            characterize and distinguish the hamate of hominoids
more palmar extension similar to that of modern                from those of platyrrhine and cercopithecoid monkeys.
humans. He argued that this provides an advantageous           Furthermore, differences between hylobatids and great
lever arm for flexion of an extended wrist and prevents        apes and even among great ape genera were captured
extension at the hamate-metacarpal fifth joint. Such an        without relying on a priori grouping methods like canon-
arrangement may be related to use of the hand for pro-         ical discriminant functions. This indicates that 3DGM is
pulsion during terrestrial quadrupedalism (Sarmiento,          a reliable tool for accurately capturing subtle shape dif-
1994; see also Patel et al., 2012).                            ferences in complex biological structures.
   As for the palmar extension of the modern human                The anthropoid primate phylogenetic tree was mapped
hamate, it should probably be interpreted as a modifica-       onto hamate shape and size spaces by reconstructing
tion for enhanced manipulation (Orr et al., 2013). Two         internal nodes using squared-change parsimony. This
intrinsic hypothenar muscles (flexor digiti minimi and         allowed us to perform permutation tests that identified
opponens digiti minimi) originate on the hamulus and           the presence of phylogenetic structure in both hamate
insert on the ulnar side of the fifth metacarpal shaft and     shape and size. This indicates that, overall, more closely
the base of the fifth proximal phalanx. In humans, these       related taxa tend to exhibit more similar hamates (prob-
muscles act in flexion and opposition of the fifth digit to    ably also due to phylogenetic clustering of similar loco-
the lateral fingers, thus facilitating the palmar cupping      motor behaviors). Although different anthropoid
and cylindrical grips important for a variety of human         taxonomic groups exhibit both different hamate shape
grips (Marzke et al., 1992, 1998; Reece, 2005; Niewoeh-        and size, allometric regressions reveal very small to non-
ner, 2006; Marzke, 2013). Our PLS results using the            significant (TIPS vs. PICs respectively) predictable cova-
whole hamate landmark configuration as a multidimen-           riation between hamate shape and size. Thus, our
sional shape variable (Fig. 8) indicate that a distally pro-   results show that differences in hamate size alone do not
truding hamulus is associated with both suspension             explain differences in hamate shape among taxa. We fur-
(negative values along PLS1) and vertical climbing (posi-      ther hypothesized about the functional evolution of the
tive values along PLS2), thus reinforcing the previously       hominoid wrist on the basis of the hamate.
stated arguments. However, the hominoid suspensory                We also used two-block partial least squares of shape
traits are not observed in Ateles, as previously reported      data and a matrix of arboreal locomotor behaviors to
(e.g., Lewis, 1989; Kivell et al., 2013) which has been        test the covariance of hamate shape and observed loco-
explained by the use of a prehensile tail during suspen-       motion without relying on the predictive models implicit
sory behaviors and their substantial amount of time            in regression analyses. This analysis revealed that,
engaging in quadrupedal locomotion (Cant et al., 2001;         among anthropoids, some aspects of hamate shape
see Table 3). Other studies have shown convergences            covary significantly with the frequency of brachiation,
between Ateles and hylobatids concentrated in the              arboreal quadrupedalism and vertical climbing. Future
shoulder, humeral shaft, and elbow (e.g., Rose, 1996;          work incorporating fossils will focus on further elucidat-
Larson, 1998; Young, 2003; Patel et al., 2013).                ing the evolutionary histories of hamate phenotypic evo-
   Although a PLS based on the PICs of shape and arbo-         lution and functional morphology.
real locomotion (not shown) was still found to be statisti-
cally significant (although accounting for a lesser
                                                                             ACKNOWLEDGEMENTS
portion of total covariance), the proximity of closely
related taxa in the PCA (Fig. 4, 5), size (Fig. 6), and PLS    We are indebted to the following researchers and cura-
plots (Fig. 8) highlights the fact that closely related taxa   tors for granting access to collections under their care:
not only look similar but they also do similar things.         Emmanuel Gilissen and Wim Wendelen, Royal Museum
Although differences in shape exist between closely            of Central Africa; Patrick Semal and Georges Lenglet,
related taxa (e.g., Pan is more similar to Pongo than to       Royal Belgian Institute of Natural Sciences; Judy Chu-
Homo), overall, large taxonomic groups (i.e., platyrrhine      pasko, Museum of Comparative Zoology; Eileen Westwig
and cercopithecoids, hylobatids and great apes) share          and Neil Duncan, American Museum of Natural History;
similar morphology (Fig. 5) and locomotion (Fig. 8).           Richard Thorington and Linda Gordon, United States
Therefore, our results on hamate shape and its covaria-        National Museum (Smithsonian). Ashley Hammond gen-
tion with locomotion support the idea that some strictly       erously provided most of the Symphalangus hamate
adaptive morphological changes (i.e., associated also          scans used in this study from the Bavarian State Zoolog-
with a change in function; Gould and Vrba, 1982) origi-        ical Collections. Soledad de Esteban-Trivigno offered
nate with phylogenetic events, strengthening the view          insightful methodological discussion. Paul O’Higgins and
that phylogeny and adaptation are part of the same com-        Kristi Lewton educated one of us (S.A.) in the applica-
plex (Polly et al., 2013 and references therein).              tion and use of PLS to coordinate shape data and matri-
                                                               ces of locomotor behaviors. A “Transmitting Science”
                                                               course on GM and phylogenetics imparted by Chris Klin-
         SUMMARY AND CONCLUSIONS                               genberg during September 2013 near Barcelona inspired
   In this work, and with a deep exploratory spirit, we        part of the analyses performed in this work. We are
sought to quantitatively characterize a biologically com-      especially grateful to Siobh
                                                                                           an Cooke and Claire Terhune
plex morphology, such is the case of the anthropoid            for inviting us to provide a manuscript to this special
hamate, using three-dimensional geometric morphomet-           issue of The Anatomical Record. This is NYCEP morpho-
rics. Major trends of shape variation were inspected in        metrics contribution number 86.
You can also read